Captions is a generative AI video creation and editing app that transforms raw footage into fully-edited, stylised videos in minutes. The platform automatically cuts scenes, overlays B-roll, adds animated captions and subtitles, and offers custom AI avatars for scalable content production. It serves creators, marketers, and businesses across web, iOS, and Android, and has become one of the better-known AI-native creative apps for short-form video.
The product is built by New York-based AI lab Mirage. In September 2025, the company rebranded the corporate entity from Captions to Mirage to reflect a broader strategy: the team is now positioned as a multimodal AI research lab building foundation models specifically tuned for short-form video on platforms like TikTok, Reels, and Shorts. The Captions app continues to operate as the company's flagship consumer-prosumer product, with the Mirage Studio brand targeting brands and advertising workflows.
Funding to date is roughly $100M, with the most recent round being a $75M growth investment led by General Catalyst's Customer Value Fund in March 2026, after earlier rounds from Index Ventures, Sequoia, and Kleiner Perkins. The company has also reportedly trained models specifically tuned for pacing, framing, and attention dynamics in short videos. In January 2025, Captions/Mirage switched to a freemium model to compete more directly with ByteDance's CapCut and Meta's Edits.
Captions' differentiation is opinionated, AI-native editing. Rather than expose every timeline knob, the product makes high-quality decisions on cuts, transitions, captions, and B-roll automatically, so non-editors can ship polished short-form video quickly. AI avatars and voice cloning extend that idea to fully synthetic talking-head content, used by marketers for repeatable explainer and ad production.
The app competes with CapCut, Adobe Premiere Rush, Descript, and a growing field of AI video tools. Its bet is that vertical integration — owning the models, the editing logic, and the consumer app — produces better short-form output than tools that bolt AI features onto traditional NLEs.